Amazon Unveils 3 New AI Agents That Can Code for Days

Amazon has introduced three new AI Agents that aim to change how teams build and secure software. The highlight is Kiro Autonomous Agent, an upgraded version of Amazon’s earlier coding tool. It learns how developers work and then handles large tasks on its own for days.

Kiro uses a concept called spec-driven development. It studies your code, your standards, and your workflow. Then it uses that learning to write new code with minimal help. AWS says the agent even tracks context across sessions and never forgets what it was assigned to do.

How Kiro Learns and Works on Its Own

Kiro gradually forms specifications while developers confirm, correct, or guide its early assumptions. It keeps these details in long-term memory, so it can tackle complex tasks without repeated instructions.

Amazon claims you can assign one big job, such as updating critical components across many internal applications. Kiro will then plan the work, execute the changes, and deliver results without constant oversight.

AWS Adds Security Agent and DevOps Agent

Amazon also revealed two more AI Agents to complete the workflow. The AWS Security Agent watches code for vulnerabilities in real time. It tests security after development and then suggests fixes automatically.

The DevOps Agent handles performance tests and compatibility checks before teams push new code live. It aims to reduce surprises during deployment and cut down errors in cloud environments.

Why These Agents Matter Now

AI Agents that run long tasks are becoming a hot trend. Other models, including OpenAI’s latest coding agent, also promise longer work cycles. Developers still face issues like hallucinations, which limit trust. However, growing context capacity is a big step in making AI Agents reliable co-workers instead of tools that need babysitting.

Amazon’s move pushes the industry toward agents that collaborate with teams, handle bigger tasks, and manage ongoing projects without losing track.